diff options
| author | Antoine Pitrou <solipsis@pitrou.net> | 2011-10-11 18:58:41 (GMT) |
|---|---|---|
| committer | Antoine Pitrou <solipsis@pitrou.net> | 2011-10-11 18:58:41 (GMT) |
| commit | e459a0877e0ca76e3af1057722a0e366395c661f (patch) | |
| tree | 6bc90d91766d28fa23af071a2b7a5a2b6d4b28e7 | |
| parent | 2c3b2302adb1bb00b6050afc30eacbc023379b93 (diff) | |
| download | cpython-e459a0877e0ca76e3af1057722a0e366395c661f.zip cpython-e459a0877e0ca76e3af1057722a0e366395c661f.tar.gz cpython-e459a0877e0ca76e3af1057722a0e366395c661f.tar.bz2 | |
Issue #13136: speed up conversion between different character widths.
| -rw-r--r-- | Objects/unicodeobject.c | 19 |
1 files changed, 14 insertions, 5 deletions
diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c index 091f964..fc42a28 100644 --- a/Objects/unicodeobject.c +++ b/Objects/unicodeobject.c @@ -177,12 +177,21 @@ extern "C" { buffer where the result characters are written to. */ #define _PyUnicode_CONVERT_BYTES(from_type, to_type, begin, end, to) \ do { \ - const from_type *iter_; to_type *to_; \ - for (iter_ = (begin), to_ = (to_type *)(to); \ - iter_ < (end); \ - ++iter_, ++to_) { \ - *to_ = (to_type)*iter_; \ + to_type *_to = (to_type *) to; \ + const from_type *_iter = (begin); \ + const from_type *_end = (end); \ + Py_ssize_t n = (_end) - (_iter); \ + const from_type *_unrolled_end = \ + _iter + (n & ~ (Py_ssize_t) 3); \ + while (_iter < (_unrolled_end)) { \ + _to[0] = (to_type) _iter[0]; \ + _to[1] = (to_type) _iter[1]; \ + _to[2] = (to_type) _iter[2]; \ + _to[3] = (to_type) _iter[3]; \ + _iter += 4; _to += 4; \ } \ + while (_iter < (_end)) \ + *_to++ = (to_type) *_iter++; \ } while (0) /* The Unicode string has been modified: reset the hash */ |